Module 3 - Digital I/O & Timing

10. Digital Inputs, Pull-Ups & Debouncing

Read buttons reliably: pull-ups and active-low wiring, contact bounce you can see on the oscilloscope, debouncing, and acting once per press.

Estimated time 2-3 hours

Learning outcomes

Explain why an input must not float and how a pull-up fixes it
Wire a button to GND with INPUT_PULLUP and read pressed as LOW
Capture contact bounce on the oscilloscope and measure how long it lasts
Debounce a button with millis() so one press is accepted once
Detect the press edge and report state changes on Serial without flooding

Parts and preparation

Uno, push button, jumper wires, onboard LED and an oscilloscope.

Before power: inspect wiring, confirm supply voltage and ensure all connected circuits share GND.

Floating inputs

An Uno input pin has a very high impedance - it draws almost no current, so nothing holds it at a level when it is unconnected. A floating input picks up noise from nearby wires and your hand, and reads HIGH or LOW at random.

Try it: set D2 to plain INPUT with a loose jumper attached, print digitalRead(2) in a fast loop, and move your finger near the wire. A pull-up or pull-down resistor fixes the default level so the pin is defined whenever the button is open.

INPUT_PULLUP

The ATmega328P has internal pull-up resistors of roughly 20-50 kΩ that pinMode(pin, INPUT_PULLUP) switches on. Wire the button from the pin to GND: open reads HIGH, pressed reads LOW. That active-low wiring is the course pattern for every push button.

Pressed current is tiny - about 5 V / 30 kΩ ≈ 0.17 mA. For long cable runs or noisy environments, add an external 10 kΩ pull-up to make the input stiffer.

Wiring	pinMode	Open reads	Pressed reads
Button to GND (course)	INPUT_PULLUP	HIGH	LOW
Button to 5 V + 10 kΩ to GND	INPUT	LOW	HIGH
Button to GND, no pull-up	INPUT	Random (floating)	LOW

See the bounce on the oscilloscope

Put the probe on D2 with the ground clip on GND. Set 1 V/div and 500 µs/div, trigger on a falling edge at about 2.5 V, and use single-shot (single sequence) mode. Press the button once.

Instead of one clean drop from 5 V to 0 V you will usually see several transitions over 0.5 to 5 ms before the line settles. Change the trigger to a rising edge and capture the release too - it often bounces as well.

Record the longest bounce you see across ten presses in your logbook, then choose a debounce time comfortably longer (at least double).

Scope setting	Value	Why
Vertical	1 V/div, DC coupling	5 V swing fills the screen
Horizontal	500 µs/div	Shows about 5 ms either side of the edge
Trigger	Falling edge, 2.5 V, single	Freezes the first press

Contact bounce

A mechanical button does not snap cleanly between open and closed. As the contacts meet, they bounce for a few milliseconds and the pin voltage chatters between HIGH and LOW.

If your sketch reacts to every raw edge, one physical press can look like many presses. That is why counts race, LEDs flicker, and menus jump.

Ideal: one clean edge. Real button: several false edges during bounce, then a stable LOW.

Software debounce

Debouncing accepts a new state only after the raw reading stays unchanged for a short time (debounceMs, often about 20-50 ms).

In this lesson's sketch, every change to the raw reading restarts a millis() timer. When the timer reaches debounceMs and the reading still differs from stableState, the sketch updates stableState. Bounce edges never last long enough to be accepted.

Restart the timer on every raw change. Only a full stable window updates stableState.

Edge detection

Acting while the button is held (level detection) fires again every loop pass. An event should happen when the stable state changes - for an INPUT_PULLUP button, that is usually the falling edge HIGH to LOW.

This sketch toggles the LED only when stableState becomes LOW, so one press equals one toggle.

Level: many actions while held. Edge: one action when stableState becomes LOW.

Report changes, not levels

Printing the button state on every loop pass floods the Serial Monitor with thousands of identical lines per second and hides the one line you care about. Print only when the stable state changes - the same edge that triggers the action.

The practical assessments mark this directly: Serial must report each state change once, with no flood while a button is held.

Wiring and safe build sequence

One button side -> D2
Opposite button side -> GND
Configure D2 as INPUT_PULLUP
Oscilloscope probe on D2, ground clip on GND

Power rule: switch off before moving wires. Arduino I/O pins are control signals; high-current loads require a driver and suitable external supply.

Worked sketch

Download .ino sketch

const byte buttonPin = 2;
const unsigned long debounceMs = 30;

bool stableState = HIGH;
bool lastReading = HIGH;
unsigned long changedAt = 0;
unsigned int pressCount = 0;
bool ledOn = false;

void setup() {
  pinMode(buttonPin, INPUT_PULLUP);
  pinMode(LED_BUILTIN, OUTPUT);
  Serial.begin(9600);
  Serial.println("Ready");
}

void loop() {
  bool reading = digitalRead(buttonPin);
  if (reading != lastReading) {
    changedAt = millis();
    lastReading = reading;
  }
  if (millis() - changedAt >= debounceMs && reading != stableState) {
    stableState = reading;
    if (stableState == LOW) {
      pressCount++;
      ledOn = !ledOn;
      digitalWrite(LED_BUILTIN, ledOn);
      Serial.print("Press ");
      Serial.println(pressCount);
    }
  }
}

How the code works

The timer restarts whenever the raw reading changes, so bounce never lasts long enough to be accepted.
Unsigned subtraction remains safe when millis rolls over.
The LED toggles and the count increases only on the stable pressed edge.
Serial prints once per press, so holding the button prints nothing extra.

Test and record evidence

Expected result: Each physical press toggles the onboard LED once and prints Press 1, Press 2 and so on. The scope shows a few milliseconds of bounce that the sketch ignores.

Practical evidence checklist

Labelled wiring or block diagram Uploaded sketch and observed result At least one measured/tested value Fault or improvement recorded in logbook

Common faults and checks

Four-leg buttons connect pairs of legs internally; rotate the button if it always reads pressed.
Count jumps by two or more per press: debounceMs is shorter than the bounce you measured - increase it.
Presses are missed when tapped quickly: debounceMs is too long; aim for about twice the measured bounce.
Nothing prints: check the Serial Monitor is at 9600 baud and the button is on D2 to GND.
Scope shows no edge: trigger level above 5 V or trigger on the wrong slope.

Extension challenge: Tell a short press from a long press: print SHORT for presses under 600 ms and LONG for presses held longer, measured with millis() from the press edge to the release edge.

Check your understanding

Q1. Why must an input not float?

Show answer

Its logic level would be undefined and noise-sensitive.

Q2. With INPUT_PULLUP and a button to GND, what does a press read?

Show answer

LOW.

Q3. What is button bounce?

Show answer

Rapid unintended transitions as mechanical contacts settle.

Q4. How do you choose a debounce time?

Show answer

Measure the longest bounce on the scope and use a time comfortably longer - at least double.

Q5. Why print only when the state changes?

Show answer

Printing every loop floods Serial and hides the events that matter.